Weak fault diagnosis of rolling bearings based on singular spectrum decomposition, optimal Lucy–Richardson deconvolution and speed transform

Abstract

Singular spectrum decomposition (SSD) has attracted considerable attention in the field of fault diagnosis in rotating machinery. However, few investigations have been carried out on its performance in processing the signals, which are composed of sub-components with time-varying frequencies. This paper attempts to introduce SSD to process the fault vibration signals of a rolling bearing under variable speed conditions. Moreover, SSD is combined with the optimal Lucy–Richardson deconvolution (OLRD) method and speed transform (ST) to construct a novel weak fault diagnosis approach for bearings under variable speed operation. Specifically, SSD is firstly introduced to separate the fault feature signal from the original signal of the bearing. An OLRD method based on Shannon entropy is then applied to the fault feature signal to further inhibit the interference and demodulate the fault information. Finally, ST is used to recognize the fault characteristic order (FCO) of the output signal of the OLRD, and the fault type is determined by comparing the recognized FCO and the theoretical FCOs of different components of the bearing. The validity of SSD for decomposing multi-component signals with time-varying frequencies is studied using a simulation signal, and the results imply that SSD is more accurate at extracting the sub-components compared to the existing algorithms, such as empirical mode decomposition and variational mode decomposition (VMD). The proposed method demonstrates good performances in detecting the fault signatures of simulated and experimental fault signals of rolling bearings under variable speed operation.